Professional Documents
Culture Documents
net/publication/226046511
CITATIONS READS
6 69
5 authors, including:
Some of the authors of this publication are also working on these related projects:
Developing an Integrated Offshore Energy Industry Safety Culture Evaluation and Improvement Toolbox View project
All content following this page was uploaded by Jerome J. Congleton on 03 December 2015.
It--
K SMU
\t lA. Ap
ax ----- -- '.
a•. %
R tc P v' FDr ) UA4ErqITA1'Kf; PAI~E
59
UNCLASSIFIED
GYNE.ALX ýNSTIU i!S OCOW'ALETING F 2993
T7he Repor Dom anm to :q- (RtOP) isued in ai nciunonq anod car loi no reporls. II is irfoloFý
that -wi van!o~ SI' Ow'i io of~'.rthe repnt Imcliuilarly the coveor and tifle page.
~~'v':
In~ra ro2~or - a Ow. 'tarn fo~1ow It is Oparlart to stay witAM the lines LO meet.
Bloc vvov q
. (hem",!
1L'>My Wok1?..Drruro/viQ~ytt' I
cualK vniauty or FF0 tllions, Cit o ny
L1~'o(,~o$o,
O w own" ^, d v ,i bl e i .1t00 nt'/q -,o r s1p'.2ao
e d Ia ma r g i nA !o o nal
U) ''' Ol or vr~~r'DOD'
r('i~~~~~~~e~
. . FiO rr eWd\ alc<0'Fk . i.$rb
! to
~Io~:~
~ . .' '~ h r-.'~t) Eock 14. ~'h~tCrisKeyvvords or ph' s,?
,iriOSell
it (J I r
.. (j 7 f
AfAL.L MR
AN EVALUATION OF
VOICE STRESS ANALYSIS TECHNIQUES
IN A SIMULATED AWACS ENVIRONMENT
A Thesis
by
WILLIAM ARCHER JONES, JR.
Acoession Fo,
NTIS GRA&1
DTIO TAB
Unannounoed Q
.7ustifioation-
August 1990 Dy
Distribution/
Availability Codes
Avail and/or
Diat Special
Major Subject: Industrial Engineering
A Thesis
by
,4IJerome J~ongleof
"(ChairorCommittee)
R. Dale Huching
(Member) c.,-
August 1990
iii
ABSTRACT
The purpose of this study was to determine if voice analysis algorithms are an
scenarios.
Two inuependent workload measures were used to identify levels of stress: 1) a
predictor model developed by the simulation author based upon scenario generated
stimulus events, and 2) the duration of communication for each weapons director,
representative of the individual's response to the induced stress. Between eight and
eleven speech samples were analyzed for each of the sixteen Air Force officers who
significant vocal indicators of stress, while amplitude shimmer showed no signs of any
significant relationship with workload or stress. Consistent with previous research, the
frequency algorithm was identified as the most reliable measure. However, the results
did not reveal a sensitive discrimination measure between levels of stress, but rather,
did distinguish between the presence or absence of stress.
The results illustrate a significant relationship between fundamental frequency
and the effects of stress and also a significant inverse relationship with jitter, though , ,
iv
Th>' less dramatic. Applied research in this area must investigate the predictive power,
within subjects, of these measurement techniques. • # . r
V
ACKNOWLEDG EMENTS
I can't begin to thank all the people who invested their time and energy into me
during my tenure here at A&M.
A hearty thanks to Dr. Jerome J. Congleton, committee chairman, for your
vision in the summer of '88 and constant support since. Thank you Dr. R. Dale
Huchingson, for the wealth of knowledge and experience you shared wiih me during
my stay . A special thanks to Dr. Schiflett and my new family at the School of
Aerospace Medicine, for taking care of me and for all the great memories. I want to
mention you all by name but I don't have five t itra pages. To Tom Doherty, thanks
for the software and patience needed to accomplish the goal. A big thanks goes out to
my Homey, Lt Slater. Tim, I couldn't have done it without youl
Eternal thanks to my family, who are, and always will be, pillars of strength
and love. Thanks Pastor Jeff, Marianne, and the Church for the love and
encouragement I've received. Between you all, I know there were prayers for me
heard in Heaven that could move 1000 mountains. Terri, thank you, just for being you
and giving me the TLC I needed to "keep the Faith".
Thank You, Jesu:,, for Your faithfu'ness and grace to guide and sustain me.
Jesus said unto him, "Thou shalt love the Lord thy God with all thy heart, and with all
thy soul, and with all thy mind. This is the flst and great commandment. And the
second is like unto it, Thou shalt love thy neighbour as thyself. On these two
commandments hang all the law and the prophets". St. Matthew 22: 37-40
vi
TABLE OF CONTENTS
Page
ABSTRACT ................................................................................ iii
APPENDIX
A. BROOKS AFB RESEARCH PROTOCOL .............................. 27
B. SUBJECT DATA ...................................................... . 40
C. SPEECH ANALYSIS SUMMARY DATA .............................. 42
D. STATISTICAL RESULTS FROM ANOVA,
REGRESSION, AND CORRELATION DATA ........................ 51
VITA .......................................................................................... 59
vii
LIST OF FIGURES
Figure Page
1. Laboratory configuration of the AWACS WY)
crewstation (C3 GW ) .......................................................... . 7
2. Graph of predictor model workload
measurement .... 4. .......... .........
. .. ..... 9
3. Graph of duration of comir nication
workload measurement .... .......... ...................... 9
4. Optional graphic display during pitch extraction
in SW IFFT .......................................................................... 10
jitter - under stress, the variability of pitch in the voice (jitter) decreases, 3) amplitude
speech rate - under stress, the rate of Epeech increases, and 7) energy distribution -
under stress, the proportion of energy between 500-1000 Hz increases (Brenner and
Shipp, 1988; Kuroda et al., 1976; Schiflett and Loikith, 1980; and Scherer, 1981).
Amplitude has only limited application in environments with consistent control over
both distance and angle of the recording device to the crewmember's mouth and thus,
may prove impractical for operational settings. Equipnment and software were not
available to include the PSE, speech rate, and energy distribution techniques, therefore,
the present study will incorporate fundamental frequency, frequency jitter, and
reveals inconclusive evidence as ., their validity (Biers, 1984). However, the nature of
the evidence must be challenged. The vocal parameters themselves may be sound, but
the adequacy of the stressors employed are in question. Previous research, conducted
in clinical settings with voice and physiological parameters of stress, has shown
positive correlation between these measures (Brenner and Branscomb, 1981; Brenner
and Shipp, 1982; and Brenner et al, 1983).
This study employed a similar approach to evaluate voice stress measures
Shipp (1988) found several significant relationships between their chosen voice
measures and artificially induced levels of stress. The present study utilized three of the
same vocal measures but extended the research beyond simple synthetic laboratory
tasks to a complex decision-making study, where situational stress was not artificially
4
produced but the result of a realistic and dynamic military mission. Manipulation of
this worldoad.*intensive simulation and direct measurement of specified tasks within the
simulation scenario provided the independent measures to effectively evaluate the
chosen voice measures of stress.
The primary objective of this study was to evaluate and comparm fundamental
techniques and two independent workload measures allowed for comparison with
previous research results and provided insight as to potential applications in voice strss
analysis technology.
5
METHODOLOGY
second objective was to assess the performance degradation of individual and team
performance produced by the C3 scenarios of high and low workload (Eddy, 1989 and
1990).
Although the primary goals of tie USAF project were drug related, the placebo
and Seldane groups were analyzed in the present study without respect to different
medication conditions. A repeated measures Analysis of Variance (ANOVA) was used
to compare Medication X Time and no significant differences were found between the
Seldane medication and placebo effects (C. Oakley, personal communication, May 1,
1990). Therefore, speech communications from both groups were used for the stress
analysis data.
Subjects
Sixteen Air Force officers (thirteen male and three female with an average age of
26) from the 5 5 2d Air Wing, Tinker Air Force Base, Oklahoma, were assigned to
Brooks Air Force Base, Texas, to spend their work week in support of the study. The
6
subjects were all weapon directors (WDs) with between 265-2000 hours of flight and
simulator time logged (See subject data in Appendix B). All participants had previously
volunteered for the study. WDs are primarily responsible for C3 missions in the
airborne environment. The E-3A AWACS aircraft is the platform from which WDs
accomplish their mission. All subjects signed the, Advisory Committee on Human
Experimentation (ACHE) approved consent form prior to any data collection and female
subjects had a pregnancy test within the previous thirty days and signed a pregnancy
disclaimer (Appendix A).
AR~aratus
_..
.........
IIIIi
the widest variety of WD performance tasks. The DCA mission consists of operations
conducted to detect, identify, intercept, and destroy enemy aircraft attempting to
penetrate friendly airspace. Each scenario was based on a standard enemy attack of
four waves. Figure 2 is a model of the predicted workload throughout a three-hour
scenario, measured in terms of WD activity units per five minute interval. Minutes 0-
15 represent the baseline level area, while the four waves are depicted by the rise in
activity units at minutes 30, 60, 90, and 120. Figure 3 is a graph of a typical individual
WD's measured duiration of communication per five minute interval,
007E
160 EEH-±--
------.---
140 "00 -----
-
.......
z 120
~100-0
F80
60
40-0
20 0
120
40
-, . . .. I . 0 I
D~G. 1TiII.V 16
-ý.Ayp.0.8460 V
0 FO~P:
- i ttor: i
L36'.97
0.0995 Hb
IZ
V.
S-Shimmojr: 362
Figure 4. Optional graphic display during pitch extraction in SWIFFT. The wave form
of each period may be shown to permit visual confirmation of correct markings of
successive cycle boundaries.
Procedure
Ex ntal Desgn
Individuals have their own unique voice print, much like fingerprints. The design
of this experiment responded to the fact that the absolute values of the WD's voice
12
parameters were of little value, when comparing between individuals. Data was
analyzed in terms of relative changes in voice throughout the simulation profile, by
utilizing within subjects analyses.
Two independent measures of stress were employed to identify times of high
workload in the scenario. The predictor model provided simulation-dtpendent
workload measures. The simulation author developed this model, based upon scenario
manipulations such as number of hostile aircraft launched, using WD activity units per
five minute interval as a measure of workload. Activity units are defined as the
summation of 1)the number of footswitch microphone activations and 2) the number of
tactical bearing and range, commit, and reinitiate button pusheg. The duration metric
provided individual WD-dependent data, based upon the crewmembers' reaction to
scenario events. The WD's duration of communication within five minute intervals
was considered an accurate measure since their primary duty is to communicate C3
information between the AWACS and pilots under their control. The three voice
measures of stress (fundamental frequency, jitter, and shimmer) were the dependent
measures.
A two-way, repeated measures analysis of variance (ANOVA) model was used
to evaluate the ability of the three voice algorithms to distinguish between five levels
within the predictor model (baseline, low, medium, high-i, and high-2). Only the
predictor model allowed for this analysis because the defined stress levels were
consistent between subjects. Duncan's multiple range test was used for specific post-
hoc comparisons between treatment means. A within subjects simple regression and
correlation analysis was also employed to determine if a relationship exists (and to what
extent) between the voice stress algorithms and each of the workload measures.
13
RESULTS
flrediQLt M~de
The relationship between the five levels of workload (defined by the predictor
model) as a function of frequency, jitter, and shimmer means is depicted in Figure 5.
160 1.3 .
-" 1.2
>. 150 0- 1,1 ... ....
*"-
100
14aaaaaaa0"-
ai- ,. .. . .... -a 0.8 . . .
130 BASE LO MED HI W2 0.7 BASE
: " LO•: MED
• HI: - H2
10.8
...10.6 .. . .
10.2
10.0
" 9.8
UM)
BASE LO MED HI H2
WORKLOAD LEVELS
Figure 5. Predictor model analysis means as a function of frequency (upper left), jitter
(upper right), and shimmer (bottom) at five levels of workload.
14
An ANOVA revealed that significant effects were obtained with the fundamental
frequency measure, E(4,15) = 7.63,11<.000l. Follow-up Duncan's multiple range
tests of the data show the baseline grouping to be significantly different from the four
other levels of stress at (pt<.01) and different from the med, hi, h2 levels at (J1.001).
No other statistically significant differences exist between the workload levels. Both
frequency jitter and amplitude shimmer parameters of voice were found to have no
significant differences (See ANOVA tables in Appendix D).
Simple linear regression analysis was run on each WD (within subjects) to
study the relationship between the three dependent variables and the predictor model
(Appendix D includes WD regression and correlation values). The results showed the
frequency measurement to be positively correlated with levels of stress in 15 of the 16
subjects (Figure 6). The pooled correlation was significant at the (1<.000) level (See
Table 1).
Table 1.
Pooled regression and correlation data.
240
220fv? loe
............
1140-
660
1o0
ACTIVITY UNITS1
Slopes
POStIOVS
2.
.. . . . . .,. ...
. .
0.
TO' 110 ;GO 1; '140 '4
AM' IVITY UNITS
Positive &lop#$
4 ,
AOTIMVY UNRti
A similar, though not as obvious, relationship was found to exist in the analysis of
jitter, with 11 of the 16 negative slopes (Figure 7). Pooled data showed a significantly
negative relationship with levels of stress (<.I05). Shimmer data revealed no
DurationQ Communicin
Total time involved in subject communication varied from 4-147 seconds within
each five minute interval. The results of the duration of communication data analysis
was similar to the model data results, but with a higher degree of confidence. The
number of speech samples were increased per WD and the fundamental frequency
measure showed a significantly (p.<.0001) positive relationship, with all slopes
following an upward trend (Figure 9). Jitter values were also shown to be
significantly correlated (p<.05), but in the negative direction, as shown in Figure 10.
Once again, the shimmer measure has shown no significant negative relationship with
the duration of communication workload metric (Figure 11).
260.
240.
ISO.
LU 140-
120.
100.
so.
Go 0. . . 2b
.. . b " " o •k " 6,o
0"" "1"
i o ' il"
DURATION OF COMMUNICATION (SiC)
.Negalive Slopesl
2 .
.21
15 ** ~ - Po$itive slopes
Negative Slopes
..............
w 15-
...............
5..
DISCUSSION
The voice stress measurement results showed a consistent pattern with both the
Shipp (1988). For example, fundamental frequency proves to be the most salient vocal
indicator of psychological stress. The frequency jitter measure responded in the
predicted fashion, but with marginal statistical significance and less confidence.
Amplitude shimmer produced insignificant results and showed no promising trends of a
whether the WD is pre-flighting hi,-. system on the ground vs flying, but not whether he
is in combat or controlling for a refueling mission. However, this does not mean that
this voice measure has no potential operadional application. Perhaps these findings are
a result of the model representations of medium, hl, and h2 being too close together.
The disparity between the medium level of activity and the h2 level is only 15 units.
The model's peaks were selected for analysis to coincide with the scenario's high
not every WD responds at the same time or in the same way to a given stimulus in the
simulation. The crewmembers work in teams of three and an individual WD's reaction
to the model's representation of high workload is a direct result of team interaction.
20
The team may be running behind on a mission and the actual stress effect may not be
confronted until five or ten minutes after the scenario stimulus was engaged; and that
loading may last throughout the initiation of the next stimulus. The fact that high task-
loading occurred in the scenario is not the question; it is whether the predictor model
identified the coirect sequence of tirae related to subject speech. It may have occurred
earlier or later or may not have been manifested in the words "Roger" or "Copy".
Regression analysis presented a positive relationship between the predictor
model and the frequency measure and a negative slope relationship with the jitter
technique. As displayed in Figures 6 and 7, the trend is extremely strong with
frequency but jitter shows a much less consistent trend. Shimmer reflects no consistent
trend in either direction as is evident in Figure 8, when a negative slope was the
expected result.
The increased number of speech samples per WD and subject specific nature of
the duration workload measure, resulted in greater statistical evidence of a positively
correlated relationship with frequency and a negatively correlated one with jitter.
Figure 9 exhibits a consistent trend with every regression line having a positive slope.
Jitter has some regression lines that are not in the expected direction (Figure 10), so
confidence of this result is not as strong. The shimmer measure shows no degree of
sensitivity to the chosen voice measures. Figure 11 shows the random spread of slopes
in both directions.
The duration of communication workload measure was more representative of
the individual effects of stress that WDs experienced during the simulation, since it was
empirically derived. Task-loading variance between subjects was taken into account in
this measure, but the adopted definition of workload could be questioned. The WD's
primary duty is vocal communication, however, during a five minute interval of high
communication, the person could conceivably become stressed and unstressed several
21
times. Decreasing the intervals of time would identify more distinct high workload
areas but finding similar utterances throughout the scenario within such small periods
would prove impossible in the present application.
22
CONCLUSIONS/RECOMMENDATIONS
conditions.
Future studies need to recognize the necessity for definitive independent stress
stress. Objective physiological data is one accurate mode of measurement but due to
the required fidelity of this AWACS simulation, all obtrusive measuring devices were
not included. Also, this study was limited to only the raw individual performance
measures collected from the AWACS data. Processed individual performance data, that
will be provided by researchers at Brooks AFB in the near future, should increase the
accuracy and precision of identifying specific workload occurrences. The need for
accurate assessment techniques of specific times that individuals are stressed, can not
be emphasized enough.
Continued research in this area is required to provide reliable measures of
workload and stress. Detailed performance data from the Brooks AFB AWACS study
should be used as independent workload measures to evaluate these and other methods
of voice stress analysis. Also physiological measurement techniques should be
investigation. Greater samples per subject, to establish a regression line may prove to
applications extend from aiding in the investigation of aircraft accidents and mishaps to
forecasting pilot disfunction caused by overload to providing filter algorithms to modify
Voice stress analysis is maturing as a research area and promises a bright future in
manned aerospace and space technologies.
25
REFERENCES
Brenner, M., Shipp, T., Doherty, T., and Morrissey, P. (1983). Yoaluare 2f
Ss Labrat= and = data. Paper presented at the
Conference on Physiology and Biophysics of Voice, University of Iowa.
(Private collection, W. Jones).
Kuroda, I., Fujiwara, 0., Okamura, N., and Utsuki, N. (1976). Method for
determining pilot stress through analysis of voice communication. Aviation.
s and environmental mediine, 47, 528-533.
Ruiz, R., Legros, C., and Guell, A. (1990). Voice analysis to predict the
psychological or physical state of a speaker. Aviation. p And
ienvironmont1
medicine, 266-271.
APPENDIX A
centrally-active antihistamines such as Benadryl (Fink & Irwin, 1979; Clarke &
Nicholson, 1978; Nicholson, Smith, & Spencer, 1982; Nicholson & Stone, 1982;
Kulshrestha, Gupta, Turner, & Wadsworth, 1978; Betts, Markman, Debenham,
Mortiboy & KcKevitt, 1984).
All of the studies cited above used simple performance tasks. The impact of the
newer terfeiiadine medication on nomplex tasks is unknown. Demonstration of an
absence of adverse effects on complex tasks under the terfenadine condition could
potentially reduce grounding time by supporting a medical flying waiver. Complex
performance tasks, such as the Complex Cogriiti e Assessment Battery, are beginning
to appear (Samet, Marshall-Mies & Albarian, 1987). They have not, however, been
normed or validated against complex, real-world work environments. At the present
time, assessing the performance effects of these medications can best be accomplished
in a simulation of real-world complex tasks under sustained operations.
Another issue of importance to this study is the assessment of drug effects on
teams solving complex problems. Although there are several models for evaluating
teams, most require inputs from trained observers maldng. subjective ratings. Reliable
detection of subtle effects of medications and fatigue require objective, repeatable
measures. Eddy (1989) and Dyer (1986) both concluded that no one has systematically
developed and empirically tested a comprehensive theory of team performance. As a
result, Eddy and Shingledecker, under contiact to the Air Force, have developed a
hierarchical performance assessment system. This system includes four levels of
metrics that can be interrelated. The levels include measures of mission effectiveiiess,
system performance, human performance, and performance capability. The upper
levels were developed in conjunction with subject matter experts (SMEs) in AWACS
C3 tasks. The system and human performance levels include several types of objective
team measures such as situational awareness, cooperation, cohesiveness, adaptation,
and distribution of work.
The proposed study will use six empirically derived, unclassified, air defense
AWACS scenarios to evaluate two antihistamine medications against a placebo using a
wide variety of performance measures. Three of the scenarios are high workload and
three are low, as verified by subject matter experts and pilot test teams.
9. Experimental Method:
a. Study Design: The proposed study will use a double-blind design with a different
drug administered to each of three groups. The three drugs include Seldane, Benadryl,
and placebo control. Teams of three subjects will be tested together under placebo and
one drug (see Figure 1)in both high- and low-workload conditions over three days (see
Figureof2).these
order Eachtreatments
team willduring
receivethea different
morning andorderearly
of workload so as to balance
evening sessions the 2).
(see Figure
Teams willdaily
shows the be randomly activities,
of testingwithout
schedule assigned It was necessary
repacement include
to one oftothese a placebo
orders. Figg 3
ensure
daytowill in thatequivalence
the performance of the three
condition during the first
groups. Accordingly, testingday
thistestin.g be single-bcnd the experimenters will
be aware drug condition
of thebeginning on .thisof theonly.
day training subjects
Theday and continuing blind to the
will remainthroughout the
drug condition the evenng
study.
30
Both simple cognitive tests (Perez, Masline, et al., 1987) and complex
laboratory tasks of the CCAB will be administered each day between the C3 scenarios.
Correlations of these test scores with those of the tasks embedded into the scenarios
will provide data to assess the feasibility of predicting complex "real-world"
performance from laboratory tasks under the same medications.
We anticipate that the performance, reserve capacity, and subjective measures
will show degradation with the Benadryl antihistamine, with fatigue (order effect), and
with high workload, when compared with placebo. We do not anticipate any
degradation of performance with the Seldane antihistamine.
b. Assignment: The 552d Air Wing will assign twelve teams of three weapons
directors (subjects may be male or f.reale), who have previously volunteered, to
Brooks AFB to spend their work week in support of this study. Teams will be
randomly assigned to a drug treatment condition and will receive one of the workload
orders shown in Figure 2. Female subjects must have a negative pregnancy test within
the previous 30 days and sign a pregnancy disclaimer.
c. Interventions: Subjects will arrive at Brooks AFB on Saturday or Sunday
evening for a preliminary briefing. Female subjects without proof of a negative
pregnancy test will be tested. Training will take place on Monday for approximately
eight hours. Teams will receive training on six simple computerized tests and two
3
complex tests over approximately four hours. They will also run a three-hour C
' no scenario to familiarize them with the simulated AWACS crewstations and
scenarios; no drugs will be administered. Subjects will ingest one Benadryl and one
Seldane placebo at 2230 or prior to going to sleep.
Teamnc will be tested in two 3 hour scenarios each da, for three days starting
on Tuesday. Figure 3 shows an event time-line of the dose administration and
experimental event schedule for each 24-hr session. Each group will ingest placebos
only during the testing schedule for Tuesday. A randomly assigned team will ingest the
recommended therapeutic dose of either Benadryl plus lactose placebo, Seldanz plus
lactose placebo, or both lactose placebo preparations starting on Tuesday evening.
Total antihistamine/ placebo ingestion for each group will consist of either eight 50mg
Benadryl and ten placebo preparations; four 60mg Seldane and fourteen placebo
preparations; or eighteen placebo preparations.
In order to keep the experiment doubly-blinded, dosing regimens for all groups
will follow diat for both Benadryl and Seldane. Benedryl and Seldane will have
different appearances, hence the concurrent schedules under all test conditions. Each
medication and its rlacebo will look identical to prevent the identification of the drug by
appearance. Therefore, each subject regardless of group will consume 18 capsules.
Before retiring on Monday at 2230, and again on Tuesday at 0600 after a
normal breakfast meal, each team member will ingest two lactose placebos (see Figure
1). Before being driven to the Aircrow Evaluation Sustained Operations Performance
(AESOP) facility, each team member will complete a sleep survey, a profile of moods
survey (POMS), a subjective fatigue rating scale, and a symptom (side effects)
Tuestionnaire. At 0630 the senior director will present a normal AWACS briefing of
te pcoming scenario. Next, the team will begin the first scenario for the day, which
may be either high- or low-workload, depending on the order of their assignment (see
Figure 2). Dosing with placebo will continue QTD throughout the first day of testing,
31
refer to Figure 3, until the fourth or evening dose, which will be either drug or placebo.
Depending on their group assignment, subjects will ingest either Seldane (60mg) plhs
Benadryl placebo, Benadry, (50rag) plus Seldane placebo, or both lactose placebos
before going to sleep at 2230 Tuesday evening.
At 0600 Wednesday, depending on their group assignment, each team member
will ingest the second dose of their drug treatment with the other placebo (or two
placebos) after their normal breakfast meal. All events of Tuesday will be repeated on
Wednesday and Thursday. They will not teke a drug or placebo Thursday at 2230.
Breakfast, lunch, and a snack will be provided as scheduled during the 24 hr
session. A fatigue/.ood/symptom questionnaire will be completed on each of those
occasions. The cognitive performance task batteries will be presented once each day
between the two scenarios (see Figure 3). The only free time will be in the evenings
and subjects will be expected to eat light, keep their blood alcohoi levels below (0.1%),
and retire by 2230,
Caffeine intake will be.restricted throughout the testing session. Decaffeinated
sodas, herbal tea, and water will be available periodically during the off-task time.
Smoking will be allowed in designated, outside areas, during off-task periods only.
Meals will be as low fat and low protein as possible to prevent the slower absorption of
drug into Lissue due to plasma protein binding.
d. Statistics: Data collected will be evaluated using an ANOVA with two
repeated measures (workload and day) on two grouping factors (drug groups and order
of workload). Measures of team performance and team processes will be evaluated
using a similiar statistical model, with the understanding that there will be fewer
degrees of freedom. Nonparametric tests will be conducted with the mood/symptom
questionnaire data. Multiple linear regression techniques will provide results
concerning the predictive power of simple task measures to complex real-world
measures under the various drug conditions.
d. Privacy: All provisions of the Federal Privacy Act, 5 USC 552a, will be
complied with.
e. Stopping Rule: The subject may at any time elect to discontinue participation.
The Medical Monitor may also stop a subject from participating or break the drug
treatment code. The Command, Control Center will have a copy of the code that can be
broken if a problem arises after normal clinic duty hours.
13. References:
Betts, T., Markman, D., Debenham, S., Mortiboy, D. and McKevitt, T. (1984).
Effects of two antihistamine drugs on actual driving performance. Biti
Medica Journal, 288, 281-282.
Boggs, P.B. (1987). Clinical Experience with Terfenadine Worldwide Trials.
Proceedings of Current Pharmacotherapy for Allergic Rhinitis and Urticaria.
h&Journal f R irato Disease, 2, 36-46.
Clarke, C.H., and Nicholson, A.N. (1978). Performance studies with antihistamines.
•1•1riihJournl.of Clnia Phraog, L,31-35.
Dyer, J.L. (1986). Annotated flib gnby nd =rtiof-lhe-a review Qf thx &Jgf
=tam I mg s it relates to lnilr. t ARI Research Note 86-18, ARI
Field Unit at Fort Benning, Georgia, Fort Benning, Georgia.
33
Eddy, D.E. (1989). Selected m prforance measures in AC Environment--An
nAk0oatnd Bibliog'aphv. Technical Report USAFSAM-TR-87-25, USAF
School of Aerospace Medicine, Brooks AFB, Texas.
Englund, E.C., Reeves, R.L., Shingledecker, C.A., Thorne, D.R., Wilson, K.P., and
Hegge, F.W. (1985). JLftied Tr-evc Qgitv Pefrac Assessment
BO.Ml (,T,-P.A,- Ik Desind-j.Ct, l of he]
k ly3, JW6D3
MILPERF Report No. 85-1, U.S. Army Research and Development
Command, Fort Detrick, Maryland.
Fink, M., and Irwin, P. (1979). CNS effects of the antihistamines diphenhydramine
and terfenadine (RMI 9918). ,hragL iajAis, 12, 35-44.
Kulshrestha, V.K., Gupta, P.P., Turner, P. and Wadsworth, J. (1978). Some clinical
pharmacological studies with terfenadine, a new antihistamine drug. British
Journal 2f Clinicl Ph~b nlgy, g, 25- 29.
Nicholson, A.N., Smith, P.A., and Sprucer, M.B. (1982). Antihistamines and visual
function: Studies on dynamic acuity and the pupillary response to light. Briish
Jounmal f Clinical Pharaog, 14, 683-690.
Nicholson, A.N., and Stone, B.M., (1986). Antihistamines: Impaired performance
and the tendency to sleep. HurM•an Journal Qf C PharmAg2g., IQ, 27-
pinicA1
32.
Perez, W.A., Masline, P.J., Ramsey, E.G., and Urban, K.E. (1987). Untfird Wi:
- performance assessment
eric gnitiy - Review o .tol'.
Technical Report AAMRL-TR-87-007, Armstrong Aerospace Medical Research
Laboratory, Wright-Patterson AFB, Ohio.
Samet, M.G., Marshall-Mies, J.C., and Albarian, G. (1987). E comlex
ggnitxv assessment a (t Finaltes n user
AAC-UM-33212, Analytical Assessments Center, CA: Los Angeles.
Deliverable under Contract MDA903-84-C-0449 to U.S. Army Research
Institute, VA: Alexandria.
Shingledecker, C.A., (1984). A J b for human Rerformance
assessment msearch. Technical Report No. AFAMRL-TR-84-07 1, Air Force
Aerospace Medical Research Laboratories, Wright-Patterson AFB, Ohio.
Sorkin, E.M., and Heel, R.C., (1985). Terfenadine: A review of its pharmacodynarnic
properties and therapeutic efficacy. Intratignal ournal 2f Current
CPcp== .
Sand Ph aoln.gy Rjvews, ADIS Press International,
Inc. Newtown, PA.
Spector, S.L. (1987). Ttflith.n.• of anthisamines. Proceedings of Current
Pharmacotherapy for Allergic Rhinitis and Urticaria. Th Journal 2f
Resimlrt 2 , 24-28.
34
FIGURE 1. Research Design.
Teams of three subjects will receive a different drug treatment, either Benadryl,
Seldane, or placebo. Each group will receive its drug treatment on the following
schedule.
Monday: Training - Placebo at 2230
Tuesday: Testing - Placebo 3 times, then Benadryl at 2230
Wednesday: Testing - Benadryl
Thursday: Testing - Benadryl
Friday: Free, unless needed to make up previous day.
Seldane ru
Monday: Training - Placebo at 2230
Tuesday: Testing . Placebo 3 times, then Seldane at 2230
Wednesday: Testing - Seldane
Thursday: Testing - Seldane
Friday: Free, unless needed to make up previous day.
Placebo ru
Monday: Training - Placebo at 2230
Tuesday: Testing . Placebo
Wednesday: Testing - Placebo
Thursday: Testing - Placebo
Friday: Free, unless needed to make up previous day.
Data collection will normally start no sooner than 1hour after drug administration.
Subjects will consume no more than six capsules daily. Benadryl subjects will
consume no more than 200 mg daily and Seldane subjects will consume no more than
120 mg daily.
35
FIGURE 2. Complete Design for a Group.
Workload Order 1
Tues Wed Thurs
Drug P D D
Morm Lo Lo Lo
Aft Hi Hi Hi
Workload Order 2
Tues Wed Thurs
Drug P D D
Morn Hi Hi Hi
Aft Lo LO Lo
NOTES:
1. P - Placebo Lo - Low-workload session
D - Drug Hi - High.workload session
2. In the Placebo Group, the drug would be replaced by placebo,
3. If there should be an equipment/software or subject problem, Friday will be
available for testing.
4. Four teams will be divided equally between the two orders.
36
FIGURE 3. Daily Testing Schedule
Time Activity
0600 Mission Planning & Breakfast
0630 Drug/Placebo, questionnaires
0700 Pre-brief
0730 Run Scenario
1100 Post-brief, Box lunch
1130 Drug/Placebo, briefing, questionnaires
1230 UTC-PAB
1330 CCAB
1430 Mission Planning & snack
1500 Drug/Placebo, questionnaires
1530 Pre-brief
1600 Run Scenario
1930 Post-brief, questionnaires
2030 Supper, free time (see notes)
2230 Drug/Placebo, but not on Thursday
NOTES:Precautions the night before: Bed early (10:30) light dinner. Keep blood
alcohol levels to below legal limits (0.1%).
Breakfast -juice, fruit, toast, water, doughnuts, decaffeinated soft drinks, herbal tea
Lunch - salad, vegetable soup, crackers, cookies, decaffeinated soft drinks, herbal tea,
fruit, no chocolate
37
Voluntary Consent Statement
1.The purpose of this study is to acquire sensitivity and reliability performance data on
realistic Air Defense scenario tasks under two antihistamines (drugs for allergies) and a
sustained operations schedule. The tasks require differing levels of thought, memory,
decision making, and motor response. The drugs are commercially available allergy or
antihistamine medications. A secondary purpose of this study is to acquire data relating
the actual performance impairment of these antihistamines on these complex, decision-
making tasks under normal, extended-day duty conditions.
2.1 will complete one full training day and three 15 1/2 hr testing sessions (see Figure
1). At the end of training on Monday, I will ingest the first dose of either Seldane
(60mg), Benadryl (50mg), or a placebo. The placebo is a non-active lactose sugar pill.
I understand that if I have lactose sensitivity or am allergic to milk, I should not take
part in this study. I understand that my knowledge or the investigators' knowledge of
my medication treatment may affect my performance. Therefore, neither I nor the
investigators will know which medication (antihistamine or placebo) I am taking. Only
the medical monitor will know that information. I understand that I will be taking no
more than thr recommended therapeutic regimen (four 50mg Benadryl pills, or two
60mg Seldane pills) or the same number of placebo pills during each 15 1/2 hr session,
As an additional protection against identifying which medication I will be given, I will
receive the same number of capsules per day (6) as the total of both drugs, although the
doses may be either all placebo or placebos plus daily doses of one medication,
3.The attached schedule shows (Figure 1) the times that I will take the medication and
when I will fill out questionnaires and perform the tasks. Breakfast and lunch will be
provided at the times shown on the schedule. I will drink only caffeine-free soft drhinks
and will smoke only in desiguated outside areas during off-task time.
Physician Date
40
APPENDIX B
SUBJECT DATA
41
PARTICIPANT INFORMATION
APPENDIX C
2OCT/WD2
FUNDAMENTAL
DURATION FREQUENCY (Hz) ITTER (%) SHIMMER (%)
50.48 125.11 0.63 8.76
98.3 134.19 0.57 16.09
91.36 130.51 0.62 12.89
98.39 163.88 0.72 11,93
107.74 170.23 0.61 4.61
115.62 166.64 0.98 7.32
46.79 138.54 0.38 4.25
WORKLOAD
10 125.11 0.63 8.76
60 142.67 1.5 4.56
130 141.32 0.67 7.13
145 163.4 0.8 11.94
135 130.51 0.84 8.61
44
7AUG/WD1
FUNDAMENTAL
DURATION FREQUENCY (Hz) JI'IT•ER (%) SHIMMER (%)
6.1 104.24 1.11 10.56
105.3 121.12 0.53 9.12
95.95 152.66 0.47 4.55
94.58 169.69 0.91 5.22
97,48 127.99 0.75 5.93
108,95 142.33 0.99 5.36
30.94 126.87 0.8 24.68
WORKLOAD
10 104.24 1.11 10.56
60 149.9 0.58 3.08
130 152.66 0.47 4.55
145 169.69 0.91 5.22
135 119.11 0.56 6.26
7AUG/WD2
FUNDAMENTAL
DURATION FREQUENCY (Hz) JITTER (%) SHIMMER (%)
8.25 136.99 1.22 12.18
62.6 170.26 0.61 14.85
89.71 164.14 0.58 11.3
61,29 178.05 0.72 15.02
76.57 148.38 0.34 9.09
59.64 168.35 1.83 15.59
11.5 148.53 1.17 17.18
WORKLOAD
10 136.99 1.22 12.18
60 155.31 1.17 13,44
130 157.84 1.14 13.0
145 168.69 1.69 13.24
135 148.38 0.34 9.09
45
10JUL/WD1
FUNDAMENTAL
DURATION FREQUENCY (Hz) JITTER (%) SHIMMER (%)
58.63 129.55 0.95 7.46
54.65 177.06 0.53 5.75
103.28 164.27 0.81 10.32
93.64 164.30 0.81 15.34
86.28 167.20 0.80 6.74
75.0 147.62 0.85 5.61
61.9 169.53 1.71 8.65
WORKLOAD
10 129.83 0.96 6.90
60 135.15 2.33 15.1
130 164.27 0.81 10.32
145 153.73 0.94 11.49
135 167.54 0.44 6.05
10JUL/WD2
FUNDAMENTAL
DURATION FREQUENCY (Hz) JITTER (%) SHIMMER (%)
11SEP/WD2
FUNDAMENTAL
DURATION FREQUENCY (Hz) JITTER (%) SHIMMER (%)
58.27 111.74 3.12 8.88
89.51 126.05 1.13 9.93
122.14 138.17 0.62 9.58
92.88 126.93 0.63 10.94
128.25 166.93 1.31 8.56
98.04 130.17 0.98 12.74
75.65 145.75 0.54 5.50
WORKLOAD
10 111.74 3.12 8.88
60 124.26 0.48 11.34
130 125.33 0.85 10.8
145 126.93 0.63 10.94
135 166.93 1.31 8.56
47
14AUG/WD1
FUNDAMENTAL
DURATION FREQUENCY (Hz) JrI'TER (%) SHIMMER (%)
4.71 186.80 0.60 11.67
41,89 238.67 0.58 5.97
61.48 199.96 1.65 5.29
43.11 227.41 0.89 8,70
111.12 214.44 0.92 6.06
98.61 242.87 0.68 6.82
51.18 222.01 0.76 5.21
WORKLOAD
10 186.80 0.60 11.67
60 238.67 0.58 5.97
130 235.94 0.46 6.44
145 227.41 0.89 8.70
135 210.93 1.14 8.79
14AUG/WD2
FUNDAMENTAL
DURATION FREQUENCY (Hz) JlTER (%) SHIMMER (%)
13.25 106.26 0.61 8.15
111.88 132.44 0.78 15.39
147.03 145.71 0.69 14.54
111.75 141.85 0.71 14.75
117.46 129.77 0.54 13.1
77.3 129.82 0.41 7.4
47.25 131.04 0.56 13.5
WORKLOAD
10 106.26 0.61 8.15
60 132.44 0.78 15.39
130 121.4. 0.53 9.32
145 141.85 0.71 14.75
135 129.77 0.54 13.1
48
16OCT/WD1
FUNDAMENTAL
DURATION FREQUENCY (Hz) JITTER (%) SHIMMER (%)
9.87 189.31 0.82 5.69
50.61 182.67 1.28 9.46
83.72 173.36 2.67 10.35
99.74 196.4 0.74 9.03
97.65 228.44 1.28 10.05
99.71 252.82 0.69 10.61
71.34 216.32 1.58 11.94
WORKLOAD
10 189.31 0.82 .569
60 193.27 1 83 9.47
130 173.36 2.67 10.35
145 178.68 0.72 '1 74
135 182.35 0,45 0i,47
16OCT/WD2
FUNDAMENTAL
DURATION FREQUENCY (H2z) JITER (%) SHIMMER (%)
53.22 120.17 1.04 14.37
49.65 106.77 0.41 11.18
48.42 109.30 1.23 11.70
72.4.0 144.14 0.90 6.12
68.19 132.32 0.63 6.37
74.37 153.06 1.53 7.04
40.4 115.77 1.45 7.69
WORKLOAD
10 120.17 1.04 14.37
60 111.6 0.87 8.75
130 109.3 1.23 11,7
145 11.9.25 1.95 12.4
135 132.32 0.63 6.37
49
18SEP/WD1
FUNDAMENTAL
DURATION FREQUENCY (Hz) JIITER (%) SHIMMER (%)
7.78 74.42 5.48 15.87
73.44 97.69 2.11 15.48
76.43 104.15 0,88 12.63
113.21 109.69 0.92 8.37
80.23 140.82 0.45 8.27
94.63 143.81 0.83 8,58
52.05 109.14 1.80 13.83
WORKLOAD
10 74.41 5.52 16.02
60 105.82 1.64 10.43
130 106.89 1.38 13,63
145 98,79 0.67 8.20
135 136.25 1.27 10.0
18SEP/WD2
FUNDAMFENTAL
DURATION FREQUENCY (Hz) JrlTER (%) SHIMMER (%)
16.25 93.31 8.97 7.26
44.59 117.46 1.64 16,61
93.66 130.62 1.58 15.47
63.29 113.28 0.53 5.85
85.68 137.87 2.37 18.6
67.20 147.85 0.99 9.24
47.77 108.26 1.60 17.31
WORKLOAD
10 93.64 7.72 10.19
60 116.94 1.69 17.95
130 113.03 1.33 16.40
145 110.73 1.30 16.26
135 143.29 1.03 14.01
50
21AUG/WD1
FUNDAMENTAL
DURATION FREQUENCY (Hz) JITTER (%) SHIMMER (%)
40.49 169.11 1.09 11.42
53.61 152.37 1.21 11.03
74.43 187.84 1.09 7.42
74.77 169.55 0.87 7.09
76.11 180.77 0.72 7.97
89.11 197.59 1.10 5.99
57.68 203.97 0.78 4.65
WORKLOAD
10 169.11 1.09 11.42
60 152.37 1.21 11.03
130 187.84 1.09 7.42
145 181.51 1.07 7.38
135 155.45 1.25 17.13
21AUG/WD2
FUNDAMENTAL
DURATION FREQUENCY (Hz) JILTER (%) SHIMMER (%)
16.94 144.41 0.64 7.63
116.26 145.89 0.61 6.61
125.84 181.61 0.71 6.54
89.36 177.79 0.47 6.29
111.79 195.02 0.99 5.43
84.08 183.18 0.28 5.05
25.84 148.93 0.57 5,29
WORKLOAD
10 144.41 0.64 7.63
60 172,04 0.35 11.13
130 164.61 0.35 3.95
145 177.79 0.47 6.29
135 158.46 0.34 10.89
51
APPENDIX D
ANOVA TABLES
MODEL.
ITls
[sorceFREQUENCY s ______
MODEL. JITTER
Source df SS MS F P
Subject 15 10.694755 .71298367 1.96 .0338
Levels 4 2.2259550 .55648875 1.53 .2040
Error 60 21.774445 .36290742
mi
MODEL. SHIMMER
Source df SS MS F P
Subject 15 447.23746 .29815831 3.24 .0006
Levels 4 15.036670 3.7591675 .41 .8021
Error 60 552.71289 9.2118815 1
DUNCAN GROUPING @
VITA